Phenotypic traits
It showed that there were significant differences in BM (
F11,122=91.42;
P<0.01), BL (
F11,122=74.02;
P<0.01), TL (
F11,122=82.91;
P<0.01), T/B (
F11,122=108.00;
P<0.01), FLL (
F11,122=16.14;
P<0.01) and HLL (
F11,122=31.46;
P<0.01) among different populations of the tree shrew (Fig 1). The DX population had the largest BM and BL, a smaller T/B and greater FLL and HLL. The HN population also had larger BM and BL, smaller TL and T/B and longer HLL. The KM population had the smallest BM and BL, but larger TL and T/B. The LY and KM populations showed similar trends. The populations with T/B greater than 1 include KM population, XC population, XY population, PM population, TC population and Leye population, where the TL is greater than the BL, while BL of other populations were smaller than TL. There were significant differences in skull indicators, including CL (
F11,122=38.77;
P<0.01), CBL (
F11,122=80.74;
P<0.01), CH (
F11,122=86.68;
P<0.01), ZB (
F11,122=169.10;
P<0.01), UTRL (
F11,122=62.98;
P<0.01) and LTRL (
F11,122=36.89;
P<0.01). Among them, the CBL, ZB, especially the length of UTRL and LTRL, DX and HN populations were significantly longer than those of other populations. The KM population had a smaller CL, CBL and CH (Fig 2).
Analysis of phenotypic trait variations
There was no effect on physical traits within the population (
P>0.05), but inter-population grouping significantly impacted them (
P<0.05), with no interaction between the two. When inter-subspecies grouping is random and intra-subspecies grouping is fixed, both significantly affect phenotypic traits (
P<0.05), indicating phenotypic differentiation among populations of the same subspecies. The variance components of phenotypic traits within and between populations (Table 2) showed that the average variance percentage for the 12 traits was 8.43% within populations, 72.11% between populations and 17.46% due to random error, indicating greater variation between populations. The VST between populations was 91.19%, with 81.27% from between populations and 9.24% from within populations, making inter-population variation the main source of phenotypic variation. The average coefficient of variation for the 12 traits among individuals was 1.44%, ranging from 0.50% to 3.39% (Table 3). The coefficients for BM, T/B, FLL, HLL and CH exceeded the average, while skull traits were more stable than physical indicators.
Principal component and cluster analysis of physical and skull traits
Based on morphological traits, a tree diagram was created with similar results (Fig 3). The clustering analysis results showed that the HN population (
T.
b.
modesta) and DX population (
T.
b.
tonquinia) were clustered together, while the other populations are clustered into one branch. The other branch is further divided into TC population, PM population branch (
T.
b.
gaoligongensis), DL population, ML population and LQ population branch (
T.
b.
chinensis). The second branch is the HK and XY populations (
T.
b.
yunalis), KM populations (
T.
b.
chinensis), XC populations (
T.
b.
chinensis) and LY populations (
T.
b.
yaoshanensis) (Fig 4). Principal component analysis (PCA) was conducted on the morphological traits of the tree shrew (Fig 5) and the results showed that PCA1 was 99.74%, PCA2 was 0.07% and PCA3 was 0.11%. DX and HN populations were separated from the other populations, while the KM population lived in the lower right corner and all other populations had a mixture.
Analysis of the correlation between morphological indicators and environmental factors
This study investigated the correlation between environmental factors and morphological traits, such as latitude, longitude, altitude, average annual temperature, atmospheric pressure and precipitation (Fig 6). The results showed that latitude was significantly correlated with all morphological traits (
P<0.05), with a positive correlation with TL, T/B and CH, as well as a negative correlation with residual morphological indicators. Longitude is not correlated with BM, T/B and CBL, but has a significant correlation with other traits (
P<0.05). Among them, it is positively correlated with FLL, HLL, ZB, UTRL and LTRL and negatively correlated with BL, TL, CL and CH. There had no correlation between altitude and BL and CL, but there is a significant correlation with other traits (
P<0.05). Among them, there is a positive correlation with TL, T/B and CH and a negative correlation with BM, FLL, HLL, CBL, ZB, UTRL and LTRL. There was no correlation between precipitation and BM, TL and CBL, but there is a significant correlation with other traits (
P<0.05). Among them, there was a positive correlation with T/B, FLL, HLL, ZB, UTRL and LTRL and a negative correlation with BL, CL and CH. The annual average temperature is only positively correlated with BM and negatively correlated with CH (
P<0.05). Atmospheric pressure was correlated with all traits (
P<0.05), with a positive correlation with BM, BL, FLL, HLL, CL, CBL, ZB, UTRL and LTRL and a negative correlation with TL, T/B and CH.
Canonical correlation analysis between morphological indicators and environmental factors
The results revealed six highly significant canonical correlation relationships (
P<0.01), with coefficients of 0.964, 0.958, 0.875, 0.690, 0.561 and 0.536. Longitude, latitude, atmospheric pressure and altitude significantly affect the morphological traits of the tree shrew, while precipitation has a lesser impact and annual average temperature has the smallest effect (Fig 7, 8).
Physical indicators and skull morphology are essential for studying interspecies and intraspecific relationships
(Tai et al., 2001; Ren et al., 2020a, 2020b). Adaptations to altitude, vegetation, latitude and climate lead to variations in body and skull structures, which are linked to dietary habits, habitat and living conditions. The study found significant phenotypic differentiation in
T.
belangeri among populations, with greater variation between populations than within them. This indicates that phenotypic variation primarily distinguishes populations, while some variation also occurs within subspecies. The results of this study highlighted that the DX population has a larger BM and BL, a smaller TL and T/B and longer FLL and HLL, which may be related to its rocky habitat, but further research is needed to verify this. The clustering analysis results showed that the HN population (
T.
b.
modesta) and the DX population (
T.
b.
tonquinia) were clustered together, while the other populations were clustered into one branch and further divided into TC population and PM population (
T.
b.
gaoligongensis), DL population, ML population and LQ population (
T.
b.
chinensis) branches; HK population, XY population (
T.
b.
chinensis), KM population (
T.
b.
chinensis), XC population (
T.
b.
chinensis) and LY population (
T.
b.
yaoshanensis) branches. Based on morphological data,
T.
b.
chenensis,
T.
b.
yunalis and
T.
b.
yaoshanensis from KM and XC were clustered together, indicating that these three subspecies had similar phenotypes, which was consistent with the results of our research group on the genetic differentiation of tree shrews
(Ren et al., 2023).
Changes in animal body size related to environmental factors like latitude, altitude and temperature can improve our understanding of their life histories
(Lou et al., 2012). Two explanations address how temperature affects body size and skull morphology. Bergmann’s law indicates that animals in warmer climates have smaller body sizes and larger surface areas for heat dissipation, while those in colder climates have larger body sizes and smaller surface areas for heat retention (
Ren et al., 2020a, 2020b;
He et al., 2023; Huang et al., 2012). Low and high-temperature treatments were applied to newborn piglets. Results showed that those raised in cold environments have a reduced surface area-to-body weight ratio, supporting Allen’s rule, which states that limb length decreases in cold climates
(Zhu et al., 2024). Most evidence for Bergmann’s law comes from the New and Old Arctic regions, mainly in temperate climates, where heat conservation mechanisms are effective
(He et al., 2023). However, tropical species do not often conform to Bergmann’s law as temperate species do (
Rodríguez et al., 2008). The results of this study indicated that most of the morphological traits of
T.
belangeri were significantly correlated with latitude, longitude, altitude and atmospheric pressure, while only CH was positively correlated with annual average temperature, which contradicts Bergmann’s law.
T.
belangeri is a species that diffuses from the tropics and is mainly distributed in tropical and subtropical regions. Temperature cannot explain the phenotypic differentiation of
T.
belangeri (Correl et al., 2016; Hendges et al., 2020). Canonical correlation analysis also showed that the annual average temperature had the smallest contribution to the phenotypic variation of
T.
belangeri, supporting the previous result. Precipitation can affect the resources of habitats, which may affect body shape changes
(Maestri et al., 2016; Hendges et al., 2020). Several omnivorous mammals (
Yomtov and Geffen, 2006), including primate populations
(Kang et al., 2024; Meloro et al., 2014), had shown a similar relationship between body size and precipitation. Most mammals feed on plants and the abundance and availability of these foods are directly related to precipitation in tropical forests
(Meloro et al., 2014; Portillo-Quintero et al., 2015). In this study, some physical indicators of
T.
belangeri were significantly correlated with precipitation and the results of canonical correlation analysis showed that precipitation had a significant contribution to phenotypic traits, suggesting that precipitation may have affected the food resources of habitat environment in
T.
belangeri (Correl et al., 2016; Maestri et al., 2016), which further affects its body size. The study also highlights altitude as a key factor in the phenotypic variation of
T.
belangeri.